The increase in memory chip prices began in the third quarter of 2026. For Demingli (001309), can the single-quarter gross margin in its 2027 first-quarter report be more than 3 percentage points higher than in the third quarter of 2026?
- GigaDevice moved the opposite way, rising from 66.57% to 70.32%.
- The fourth-largest customer at 65.25%, the same sentence.
- To exceed 35.89%, x needs to be about 4.7 percentage points higher than k.
The increase in memory chip prices began in the third quarter of 2026. For Demingli (001309), can the single-quarter gross margin in its 2027 first-quarter report be more than 3 percentage points higher than in the third quarter of 2026?
存储颗粒涨价是从2026年三季度开始的，德明利(001309)2027年一季报的单季毛利率，能比2026年三季度高出3个百分点以上吗？
The desk answersverifiedWalnut, from public filings
In the third quarter of 2026, Demingli's single-quarter gross margin was 32.89%, 5.95 percentage points lower than the second quarter. To reach 35.89% in the first quarter of 2027, the selling price increase would need to exceed the increase in unit inventory cost by 5 percentage points. Based on the numbers currently visible, I lean toward it not being achievable, with the neutral scenario falling between 29% and 33%.
The starting point is misremembered: the price increase did not begin in the third quarter of 2026. In the company's September 15 reply to the inquiry letter, it stated: since the third quarter of 2025, with the explosion of AI demand, storage product prices have shown exponential growth. The single-quarter data also matches: 2025Q3 gross margin 10.60%, 2025Q4 jumped to 27.14%, 2026Q1 reached 57.42%. The third quarter of 2026 is not the starting point; it is the point when the inventory dividend begins to settle. The question asked is "two quarters after the price increase," but it is actually "one year after the price increase."
Where the cheapness lies
The chain has only four steps: wafer prices rise, module makers' procurement costs follow, costs are first parked in the inventory account, then carried into operating costs with a lag based on a 1-to-3-month production cycle, and finally land in the single-quarter gross margin.
Demingli's 57.42% in 2026Q1 was not earned from selling at high prices, but from selling products made with materials purchased in 2025. The company itself admitted this in the top five customer details: the largest customer in the first half had a gross margin of 68.95%, because "the products sold were mainly produced from raw materials purchased by the company in earlier periods"; the fourth-largest customer at 65.25%, the same sentence.
In the same table, two enterprise-level customers had gross margins of 18.11% and 21.93%, because "the negotiation cycle is relatively long, there is a time lag between product price determination and production delivery, and cost increases during the period put pressure on gross margin." In the same half-year, customer gross margins ranged from 18% to 69%; the difference was not the product, but which month the raw materials were purchased.
Where the third quarter fell
The cheap wafers ran out. On June 30, inventory was 214.12 yi yuan, with raw materials accounting for 48.36%, semi-finished goods 33.42%, commissioned processing materials 10.10%, and finished goods only 5.03%—the books were almost entirely materials purchased after the price increase. On September 30, inventory was 289.86 yi yuan; cash paid for "purchases of goods and acceptance of labor services" in the first three quarters was 399.66 yi yuan, while operating costs for the same period were only 167.56 yi yuan. The money turned into materials, not gross profit.
The enterprise-level share also rose. In the first three quarters, enterprise-level revenue was 82.37 yi yuan, accounting for 28.8% of revenue, with long-term contracts and long negotiation cycles—prices set first, costs rising later, naturally eating into gross margin. Impairment also did not interfere this quarter: inventory write-downs were only 0.63 yi yuan, and 32.89% is a clean number.
Putting peers together
| Single-quarter gross margin | 2026Q1 | 2026Q2 | 2026Q3 |
|---|---|---|---|
| Demingli | 57.42 | 38.84 | 32.89 |
| Jiangbolong | 55.53 | 61.26 | 44.74 |
| Biwin Storage | 53.30 | 55.98 | 42.80 |
| GigaDevice | 57.08 | 66.57 | 70.32 |
The three module makers all pulled back together, but at different paces. Demingli fell 24.5 points over two quarters, the most among the three; looking only at Q3 quarter-on-quarter, it fell just 5.95 points, actually the least. It peaked one quarter earlier than peers and also headed down one quarter earlier; Jiangbolong and Baiwei are still around 44% and 43%, with cheap inventory not yet fully digested.
GigaDevice moved the opposite way, rising from 66.57% to 70.32%. The money from price increases did not disappear; it shifted from the module segment to the wafer segment. "Wafer price increases" are not a one-way benefit for module makers.
The first quarter of 2027 can be calculated
Taking the third quarter as the baseline, let the selling price be 1 and unit cost be 0.671. Let k be the increase in unit inventory cost relative to the cost already carried in the third quarter, and x be the selling price increase. The 2027Q1 gross margin equals 1−0.671×(1+k)/(1+x). To exceed 35.89%, x needs to be about 4.7 percentage points higher than k.
| Inventory cost k | Selling price x | 2027Q1 gross margin |
|---|---|---|
| 0% | 0% | 32.9 |
| 0% | +5% | 36.1 |
| +5% | +5% | 32.9 |
| +5% | +10% | 36.0 |
| +10% | 0% | 26.2 |
| +5% | −5% | 25.8 |
The cells that can cross the line are all ones where the selling price increase exceeds the cost increase by more than 5 points. What is needed is not "prices staying high," but "another wave." Demingli has already sold through its cheap materials one quarter earlier than peers, and 32.89% is closest to "current-period cost corresponding to current-period selling price"—this shows both that its downside is smaller than peers' and that there is not much inventory dividend left to release.
Jiangbolong's exact words on the October 29 conference call were "storage wafer prices are at a high level," "the supply-demand relationship in the storage industry remains relatively tight," and "overseas original manufacturers are still adopting a relatively cautious capital expenditure strategy in NAND Flash." The three statements point to prices staying flat, not rising another 10% quarter-on-quarter.
An easily overlooked difference
Demingli's contract liabilities at the end of September were 4.05 yi yuan, down 35.06% from the beginning of the year; Jiangbolong's were 22.86 yi yuan, up 544.74% from the beginning of the year. Jiangbolong's explanation is that "a considerable portion of inventory has been locked up by globally renowned customers through prepayments."
Demingli is carrying this 289.86 yi yuan of inventory itself: short-term borrowings 101.50 yi yuan, long-term borrowings 91.00 yi yuan, and financial expenses in the first three quarters of 5.56 yi yuan. Based on third-quarter single-quarter operating costs of 78.15 yi yuan, that is enough to sell for three and a half quarters. When prices stay flat, the risk is its own, not the customers'. This is also why it dares to use "a relatively flexible pricing strategy" to win enterprise-level customers.
Under what circumstances would I be wrong: from Q4 2026 to Q1 2027, if module prices rise a further 10% or more quarter-on-quarter while inventory costs do not move, Q1 gross margin would reach 36% to 39%; or if enterprise-grade long-term contracts are re-signed at new prices, raising old contracts of 18% to 22% to above 30%, the threshold would also be crossed ahead of schedule. Conversely, if the 2026 annual report takes an inventory write-down of 10 yi yuan or more in one go, indicating that cost has already exceeded net realizable value, then Q1 will only be lower.
Sources: Demingli, "2026 Third Quarter Report," October 26, 2026, Announcement No. 2026-093; Demingli, "2026 First Three Quarters Earnings Forecast," October 17, 2026, Announcement No. 2026-092; Demingli, "Announcement on Provision for Impairment Losses for the First Three Quarters of 2026," October 30, 2026, Announcement No. 2026-095; Demingli, "Reply to the Review Inquiry Letter on the Issuance of Shares to Specific Targets (Revised Draft)," September 15, 2026; Longsys, "2026 Third Quarter Report," October 30, 2026, Announcement No. 2026-119; Longsys, "Investor Relations Activity Record Form," No. 2026-013, November 2, 2026; Biwin Storage, "2026 Third Quarter Report," October 29, 2026; GigaDevice, "2026 Third Quarter Report," October 30, 2026.
2026年三季度，德明利单季毛利率32.89%，比二季度低5.95个百分点。2027年一季度要站上35.89%，等于要求售价涨幅比库存单位成本的涨幅多出5个百分点。按现在能看到的数，我倾向于到不了，中性情形落在29%到33%之间。
起点就记错了：涨价不是从2026年三季度开始的。公司在9月15日的问询函回复里写着：2025年三季度以来，随着AI需求爆发，存储产品价格呈现指数级增长。单季数据也对得上：2025Q3毛利率10.60%，2025Q4跳到27.14%，2026Q1到57.42%。2026年三季度不是起点，是库存红利开始结清的时点。问的是"涨价后两个季度"，实际是"涨价一年后"。
便宜在哪一环
链条只有四步：颗粒涨价，模组厂采购成本跟着涨，成本先压在存货科目里，再按1到3个月的生产周期滞后结转进营业成本，最后落到单季毛利率。
德明利2026Q1那57.42%，赚的不是卖得贵，是卖的东西用的是2025年买的料。公司在前五大客户明细里自己承认了：上半年第一大客户毛利率68.95%，原因是"销售产品主要由公司前期采购的原材料生产"；第四大客户65.25%，同一句话。
同一张表里，两个企业级客户毛利率是18.11%和21.93%，原因是"议价周期较长，产品售价确定与生产交付之间存在时间差，期间成本上升导致毛利率承压"。同一个半年，客户毛利率从18%到69%，差别不在产品，在原材料是哪个月买的。
三季度掉在哪
便宜晶圆用完了。6月30日存货214.12亿元，原材料占48.36%、半成品33.42%、委托加工物资10.10%，库存商品只有5.03%——账上几乎全是涨价后买进来的料。9月30日存货289.86亿元，前三季度"购买商品、接受劳务支付的现金"399.66亿元，同期营业成本只有167.56亿元，钱变成了料，没变成毛利。
企业级占比也上来了。前三季度企业级收入82.37亿元，占营收28.8%，长约、长议价周期，售价先定、成本后涨，天然吃毛利。这一季减值也没来搅局：存货跌价只计提0.63亿元，32.89%是干净数字。
同行放在一起看
| 单季毛利率 | 2026Q1 | 2026Q2 | 2026Q3 |
|---|---|---|---|
| 德明利 | 57.42 | 38.84 | 32.89 |
| 江波龙 | 55.53 | 61.26 | 44.74 |
| 佰维存储 | 53.30 | 55.98 | 42.80 |
| 兆易创新 | 57.08 | 66.57 | 70.32 |
三家模组厂一起回落，但节奏不一样。德明利两个季度掉24.5个点，是三家最多的；单看Q3环比只掉5.95个点，反而最少。它比同行早一个季度见顶、也早一个季度往下走，江波龙和佰维还在44%、43%附近，便宜库存没消化完。
兆易创新反着走，从66.57%升到70.32%。涨价的钱没有消失，是从模组环节挪到了颗粒环节。"颗粒涨价"对模组厂不是单向利好。
2027年一季度的账可以算
把三季度当基准，售价记为1，单位成本0.671。设k为库存单位成本相对三季度已结转成本的涨幅，x为售价涨幅，2027Q1毛利率等于 1−0.671×(1+k)/(1+x)。要过35.89%，需要x比k高4.7个百分点左右。
| 库存成本k | 售价x | 2027Q1毛利率 |
|---|---|---|
| 0% | 0% | 32.9 |
| 0% | +5% | 36.1 |
| +5% | +5% | 32.9 |
| +5% | +10% | 36.0 |
| +10% | 0% | 26.2 |
| +5% | −5% | 25.8 |
能过线的格子，都是售价涨幅比成本涨幅高出5个点以上。需要的不是"价格维持高位"，是"再来一波"。德明利已经比同行早一个季度把便宜料卖完，32.89%离"当期成本对应当期售价"最近——这既说明它下行空间比同行小，也意味着没有多少库存红利可以再释放。
江波龙10月29日电话会的原话是"存储晶圆价格处于高位"、"存储产业供需关系仍维持偏紧态势"、"海外原厂在NAND Flash领域仍采取相对审慎的资本开支策略"。三句指向的是价格横住，不是环比再涨10%。
一个容易漏掉的差别
德明利9月末合同负债4.05亿元，比年初降35.06%；江波龙22.86亿元，比年初增544.74%，江波龙的说法是"已有相当部分库存被全球知名客户通过预付款方式锁定"。
德明利这289.86亿元存货是自己扛的：短期借款101.50亿元、长期借款91.00亿元，前三季度财务费用5.56亿元。按三季度单季营业成本78.15亿元算，够卖三个半季度。价格横住时，风险是它自己的，不是客户的。这也是它敢用"相对灵活的价格策略"抢企业级客户的原因。
什么情况下我错了：2026年四季度到2027年一季度，如果模组价格环比再涨10%以上而库存成本没动，一季度毛利率会到36%到39%；或者企业级长约按新价重签、把18%到22%的老合同抬到30%以上，也会提前过线。反过来，如果2026年报存货跌价一次提到10亿元以上，说明成本已高过可变现净值，那一季度只会更低。
- 01德明利《2026年第三季度报告》2026年10月26日 公告编号2026-093
- 02德明利《2026年前三季度业绩预告》2026年10月17日 公告编号2026-092
- 03德明利《关于2026年前三季度计提减值损失的公告》2026年10月30日 公告编号2026-095
- 04德明利《向特定对象发行股票审核问询函的回复（修订稿）》2026年9月15日
- 05江波龙《2026年第三季度报告》2026年10月30日 公告编号2026-119
- 06江波龙《投资者关系活动记录表》编号2026-013 2026年11月2日
- 07佰维存储《2026年第三季度报告》2026年10月29日
- 08兆易创新《2026年第三季度报告》2026年10月30日
## 一、8.3%比32.89%更值得看
单季拆开：三季度收入比二季度多做22.73亿，毛利额只多了1.89亿，多卖的那部分只贡献8.3%的边际毛利率。二季度更难看：收入多18.34亿，毛利额反而少6.88亿，边际毛利率负37.5%。
| 单季 | 收入(亿) | 毛利额(亿) | 边际毛利率(%) |
|---|---|---|---|
| 2026Q1 | 75.38 | 43.29 | — |
| 2026Q2 | 93.72 | 36.40 | (37.5) |
| 2026Q3 | 116.45 | 38.30 | 8.3 |
这两个数连起来是一句话：德明利今年新增的收入，都是用毛利换的。二季度是赔本换，三季度刚转正到8.3%，只有存量业务的四分之一。
为什么这件事决定2027年一季度？一季度的增量订单用的全是2026年四季度以后买的高价料，只要采购价还在涨，边际毛利率就还在8%上下打转；存量能不能修复是另一回事。靠多卖把综合毛利率抬3个点，数学上先不成立——卖得越多，高毛利存量摊得越薄。
## 二、成本率67.11%是硬底
真正锁死一季度的是成本率。单季营业成本除以单季收入，一季度42.58%，二季度61.16%，三季度67.11%，两个季度爬了24.5个百分点。爬回去也要两个季度，这期间颗粒价格还在涨。
```chart
type: line
title: 单季营业成本率：从72.86%砸到42.58%，再爬回67.11%
unit: %
labels: 2025Q4, 2026Q1, 2026Q2, 2026Q3
values: 72.86, 42.58, 61.16, 67.11
```
一季度毛利率要过35.89%，成本率必须回落到64.11%，比三季度低3个百分点。三季度的成本率降不下去：9月末存货289.86亿，单季营业成本78.15亿，料够卖3.71个季度。2027年一季度结转的成本，大部分2026年9月就躺在仓库里，买价就是那个买价；四季度补的料更贵。
两个门槛。四季度采购价持平、一季度结转全是老料，售价要涨4.7%才够；四季度再涨5%、一半新料进一季度成本，要涨7.3%才够。涨价幅度得比成本涨幅多出5到7个点，这是"再来一波"的要求。江波龙10月29日电话会用的词是"当前存储晶圆价格处于高位"——高位，不是上行。
## 三、四家放一张表，这不是行业性回落
| 2026Q3单季 | 毛利率(%) | 收入环比(%) | 单季经营现金流(亿) | 存货/单季成本(季) |
|---|---|---|---|---|
| 德明利 | 32.89 | +24.2 | (69.6) | 3.71 |
| 江波龙 | 44.74 | +3.2 | +0.9 | 3.52 |
| 佰维存储 | 42.80 | (29.9) | (37.8) | — |
| 兆易创新 | 70.32 | +18.3 | +40.8 | — |
毛利率和收入环比并排，四家站在一条"以价换量"曲线上。德明利让价最狠，换来24.2%放量；江波龙基本不让，量增3.2%、毛利率守在44.74%；佰维两头不占，量和毛利率一起掉。佰维收入环比负29.9%，说明模组环节没有共同的放量红利——德明利那24.2%不是行业给的，是价格换的。
兆易创新在曲线另一端：毛利率70.32%，收入还增18.3%，经营现金流单季正40.8亿。它在颗粒端。涨价的钱没消失，是从模组搬到了颗粒——德明利一头对着客户，一头对着原厂，两头都被挤。
## 四、传导链末端不是毛利率，是现金流
公司9月15日问询函回复里把机制写清楚了："公司存储产品的生产周期通常为1-3个月"，"由于采购、生产、销售周期的间隔影响，产品成本的变化通常滞后于销售价格的变化"，上行周期里"营业成本的上涨通常滞后于销售单价的上涨"。
链条四步：原厂合约价上行，模组厂按账期或预付采购，晶圆进存货，经1到3个月生产周期结转进营业成本。2027年一季度的营业成本，对应的是2026年10月到12月的采购单。这条链落在"营业成本"，不落"资产减值"——前者跟着采购价走，后者跟着可变现净值走。
毛利率之后还有两站。第一站财务费用，三季度单季2.92亿，占收入2.51%，借钱买料，利息直接啃净利润。第二站经营现金流，三季度单季负69.55亿。毛利是账面数，现金是真金白银，两者反着走。毛利率只是链条中段的读数，末端是它在用负债表承接这轮涨价。
## 五、同一件事，两家的写法不一样
德明利三季报把存货增长310.66%解释成"公司企业级存储业务规模及在手订单量快速增长，存货规模增加主要系公司为企业级存储业务备货所致"。江波龙三季报把存货增长143.67%只写了一句"主要系采购存货增加所致"，不提订单。
差别不在措辞，在配套数字。德明利9月末合同负债4.05亿，占存货1.4%，比二季度末的8.93亿降了一半多；江波龙22.86亿，占存货8.0%，比二季度末增71.9%。德明利预付款1.51亿，同比降49.09%；江波龙22.10亿，增236.10%。
意思很直接：德明利说在手订单快速增长，可这些订单没有预付款背书，上游也没预付锁料。备货信用它自己扛——有息负债192.50亿，占总资产47.9%，江波龙是28.4%。订单质量差一个档次，表里能读出来。
也正因为自己扛，它才敢用"相对灵活的价格策略"撬企业级大客户，原话是"为了拓展企业级存储大型客户，采取了相对灵活的价格策略"。低价换份额是主动选择，代价就是那张8.3%的表。
## 六、反方最强的那条
反方会说：8.3%不是趋势，是战略性低价导入。那批订单拿低价砸开企业级客户的门，等导入完成、2027年一季度按新价续签，结构性稀释就反转，毛利率不光过35.89%，还能更高。有先例——固态硬盘2025年全年毛利率12.69%，四季度行业复苏加上客户导入顺利，全年被拉了回来。
要成立，得同时满足三件事：客户导入一季度完成、新合同价格上修、消费端不掉量。第三件最脆弱。江波龙在同一场电话会上说，"消费电子领域中，存储价格对部分客户的采购意愿形成一定压制，但高端客户、品牌客户及海外客户的需求仍保持较强韧性"。翻译一下：涨价已经在杀需求，杀的还是中低端。顺价能力有天花板，天花板就是下游的承受力。
这是唯一能改变结论的一条，但它要求"价格再涨"和"结构升级"同时发生，概率不高。维持到不了。
## 七、什么情况下我错了
不用等一季报，四个信号哪个先出，判断都得改。
2027年4月披露的2026年报，四季度单季毛利率如果在36%以上，说明成本率三季度就见顶了，一季度过线概率上升。反过来，年报存货跌价计提超过5亿，说明9月那批料的买价已经高过可变现净值，这条判断不是错了，是更确定了。
2027年一季度如果收入掉到100亿以下、毛利率反而过了35.89%，那是缩量保价，说明它宁可让份额也不降价——"以价换量"的前提不成立，整篇要重写。
前置信号是合同负债能否回到10亿以上：回得去，客户开始预付锁价；回不去，订单还是没背书。
来源：德明利《2026年第三季度报告》2026年10月26日 公告编号2026-093；德明利《关于2026年前三季度计提减值损失的公告》2026年10月30日 公告编号2026-095；德明利《关于深圳市德明利技术股份有限公司申请向特定对象发行股票的审核问询函的回复（修订稿）》2026年9月15日；江波龙《2026年第三季度报告》2026年10月30日 公告编号2026-119；江波龙《投资者关系活动记录表》编号2026-013 2026年11月2日；佰维存储《2026年第三季度报告》2026年10月29日；兆易创新《2026年第三季度报告》2026年10月30日。